Abstract: The robust Sigma-point filtering is investigated and applied to relative navigation for unmanned aerial vehicles (UAVs) in formation flight. Making use of Huber-based estimation method, the measurement update of Sigma-point filtering is achieved by solving the linear regression problem. The new Sigma-point filtering which is a combined minimum L₁ and L₂ norm estimator exhibits robustness with respect to deviations from the commonly assumed Gaussian error probability. The relative inertial equation and line of sight measurement are provided. The robust Sigma-point filtering is used to fuse relative inertial information and line of sight mea-surements, estimating the relative attitude, velocity and position. Simulation results show robust Sigma-point filtering is able to achieve more accuracy estimation than standard Sigma-point filtering and standard extended Kalman filtering.

Key words: navigation systems, vision, robustness, Kalman filter, Gaussian distribution